How accurate are Value-at-Risk models at Canadian commercial banks?

Value-at-Risk (VaR), a measure of the dollar amount of the potential loss from adverse market moves, has become a standard benchmark for measuring financial risk. In most developed countries, commercial banks are required by regulators to compute their VaR on a daily basis. VaR estimates serve as a major determinant of the banks’ capital requirement. In our study, we evaluate the accuracy of the VaR models of the six largest Canadian commercial banks. We provide evidence that the current models used by these banks for their VaR estimation are excessively conservative. An implication of this systematic overstatement is that the required capital for these banks is larger than what it should be, which is costly for these banks. We propose alternative models for computing VaR, and we show that, unlike the current VaR models used by the banks, our models are not rejected by the data.